Insurance pricing sophistication is unrecognizable from the days less than 30 years ago of mainframe computers, manual spreadsheets, paper reports and “snail mail” market intelligence requests. The race for competitive advantage is not easing off and will increasingly require a dynamic analytics approach in a continuously changing market.
The phrase “It’s a marathon, not a sprint” is regularly used to describe change processes in business. Often, we would suggest, that it is an appropriate metaphor.
But, not so in insurance pricing.
Instead, we would describe the development of pricing capabilities, particularly in the property & casualty (P&C) market as “a sustained sprint, not a marathon.” What is more, it’s a race for competitive advantage with no immediate end in sight owing to factors such as changing customer buying behaviors and the trend toward digitalization of insurance.
What is more, it’s a race for competitive advantage with no immediate end in sight owing to factors such as changing customer buying behaviors and the trend toward digitalization of insurance.
Companies at the front of this race are increasingly investing in advanced analytics and machine learning to bring speed, consistency, agility and flexibility to their pricing function. Insurers aiming to join them, and improve financial performance in the way that early adopters have,1 are likely to need to enhance capabilities in three key areas to develop a dynamic pricing cycle (Figure 1):
Let’s look at each of those three areas in turn.
Data is absolutely the foundation of pricing sophistication. All of our experience in working across P&C business lines and life and health insurance shows that more and better data yields the biggest pricing benefits compared with, say, developing new algorithms or upgrading technology.
And there is certainly no shortage of data out there for insurers to use. First, an often overlooked but valuable source is the internal unstructured data that companies already hold on policyholders. Second, the range of external data that can be useful to an insurer, such as customer behavior data from the internet, is vast and growing all the time. The broader challenge is being able to capture and use data, such as converting unstructured underwriting and claims information into a structured form and knowing which data sources are most likely to be predictive of future claims. Similarly, a third essential data component, competitor intelligence, can vary in sophistication from simple agent feedback to full-blown quantitative analysis of competitor rates.
The question then is what is stopping more insurers from becoming more data-driven? The answer is that possibility and current feasibility are not always the same thing. Many insurers have some key challenges to address if they are to make the progress of the current industry leaders in data capture, manipulation and utilization. These include:
Many insurers have some key challenges to address if they are to make the progress of the current industry leaders in data capture, manipulation and utilization.
Another trio of factors is fundamental to harnessing what advanced analytics can bring to an insurer’s pricing.
First is the sophistication of the models being used. A recent poll conducted during a Willis Towers Watson webinar confirmed once more that generalized linear models remain the industry standard but also showed a growing tendency toward the use of decision tree machine learning techniques, notably gradient boosted machines (GBMs). Vendor tools, such as our own Radar, and open source techniques continue to push the boundaries of what models can deliver.
Often, the choice of model form involves a trade-off between simplicity and sophistication. While simpler equation-based models can be less predictive, the results will generally be easier to explain to a wider range of key stakeholders. Balanced against that, those wanting to employ more sophisticated models must weigh up possible complexity in communicating results with the potential for greater predictiveness. Even when traditional model forms are used for model deployment, advanced analytics can be incorporated into the pricing process, such as using GBMs to streamline factor selection.
A second vital aspect of advanced analytics capability is the talent to develop it. Increasingly, this requires a range of skill sets — actuarial, data science and business analysis (Figure 2) — and successful collaboration between these disciplines. Such collaboration is also a route to break down the traditional silos that have existed between functions such as underwriting, claims, actuarial and finance in insurance.
Both the models and the people involved feed in to the third, and perhaps defining, aspect of what insurers can get back from advanced analytics. That is a decision support environment that brings together all models and makes their outputs available to all the people who need them, at the time they need them. This way, ownership of models and decisions doesn’t reside with one group — say, models with actuaries or data scientists — but are available to all so that pricing decisions can be taken quickly and efficiently by relevant people in the value chain.
This support can take many forms. It may be detailed scenario testing of potential rate changes to assess their potential impact. It could be feeding in expected outcomes of pricing strategy into business planning over a given time horizon and identifying accompanying issues and opportunities. For day-to-day management of the business, more advanced analytics can underpin a wide range of management information (MI) dashboards that explore issues such as progress against key performance indicators and segmentation and distribution analysis.
Technology is the enabler of a dynamic pricing environment, greasing the wheels of insurance data and analytics capability in three principal ways:
For the most part, we would describe the industry as at a “platform upgrade” stage, meaning that companies have taken steps to introduce rules-based pricing solutions, enrich structured data capture, reduce rekeying, automate some basic processes and enhance system connectivity through technologies such as automated programming interfaces (APIs).
Relatively few insurers have evolved to an “automation” level, whereby they employ sophisticated decision engines (including in some cases real-time price delivery), they fully leverage a range of data sources and they carry out complex process automation.
The difference is likely to become more significant over time.
Looking forward, for example, the industry will have an almost endless supply of data capture opportunities as the universe of connected “things” continues to expand exponentially. This is where we expect much of the future competitive race to take place.
Looking forward, for example, the industry will have an almost endless supply of data capture opportunities as the universe of connected “things” continues to expand exponentially.
Allied to this data explosion, effective rate deployment will increasingly mean a matter of hours from MI to decision to execution. The role of technology will therefore be to provide the level of agility, flexibility and speed to market that can handle greater data volumes, fits a shorter execution time frame, supports the associated need to integrate with policy administration systems and underpins multichannel distribution.
An almost inevitable consequence of such requirements will be the need for higher levels of automation. That doesn’t signal robots becoming the underwriters, claims handlers and actuaries; it means those specialists should have less manual, repetitive and time-consuming tasks to do so that they can spend more time on more valuable and complex ones.
Ultimately, of course, the reason that more insurers are investing in a more dynamic pricing capability is they see that the drivers and baseline for the way insurers go about pricing in 30 years — or just as likely three — will be very different, just as the methods of 30 years ago are a relic of the past. Progress is accelerating at an ever-increasing pace, and computing power increases in relation to Moore’s law.
Things are already moving fast, and there are opportunities for competitive advantage to be grabbed.
Now is a good time to be preparing and conditioning your pricing function to be ready for the sustained sprint that taking such opportunities and keeping pace with the market is likely to entail.
1 Advanced analytics: Are insurers living the dream? Willis Towers Watson 2019/2020 P&C Insurance Advanced Analytics Survey Report (North America), page 4, https://www.willistowerswatson.com/en-US/Insights/2020/01/advanced-analytics-are-insurers-living-the-dream-2019-2020-P-C-insurance-advanced-analytics
